INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Many studies of prefrontal (PF) cortex have concentrated on studying the response of neurons to single spots flashed in the dark. Under these circumstances neurons respond to stimuli flashed in their receptive fields (Boch and Goldberg 1989; Funahashi et al. 1990; Mikami et al. 1982; Suzuki and Azuma 1983). The visual response of PF neurons can be modulated according to the behavioral significance of the stimulus (Boussaoud and Wise 1993; Hasegawa et al. 1998b; Kim and Shadlen 1999; Rainer et al. 1998; Sakagami and Niki 1994; Watanabe 1986), for example when the stimulus in the receptive field is the match in a match-to-sample task (Hasegawa et al. 1998b; Rainer et al. 1998; Rao et al. 1997). However, in real life primates choose the targets for saccadic eye movements from complex visual scenes that contain many distracters as well as targets. In these experiments we analyzed the response of neurons in PF cortex to stimuli that served either as targets or distracters in a visual search task. We show that a subset of PF neurons can distinguish between targets and distractors and therefore are specialized for saccade target selection. A preliminary report of this study has been presented elsewhere (Hasegawa et al. 1998a).

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Two male rhesus monkeys (Macaca mulatta) were trained to perform a delayed visual search (DVS) task (Fig. 1A). In this task the monkeys had to fixate a central spot for 1 s (fixation period), after which an array of 6 small (5° diam) circular gratings appeared at symmetric locations 20° around the fixation point and remained lit for 0.5 or 1 s (cue period). Five of the gratings had identical spatial frequencies, fine or coarse, and the remaining had a different spatial frequency. After the gratings disappeared, the monkey had to hold fixation for another 0.8-1.5 s of variable delay period, and then the fixation spot disappeared and the monkey had to make a saccade to the spatial location of the unique stimulus (go period, <1 s) to earn a drop of water. The uniqueness of the target spatial frequency, not the absolute value of the spatial frequency, was critical. A given stimulus could serve as target or distractor, depending on the frequency of the remaining stimuli. The target and distractor features (target position and spatial frequency) varied randomly across trials. For some neurons that were active during DVS trials, we also tested an oculomotor delayed response (ODR) task (Funahashi et al. 1989), in which the same set of objects and locations of the target were used as in the DVS task but presented alone. DVS and ODR tasks were run in separate blocks that contained 100-300 trials. We checked the consistency of waveform of spike and frequency of baseline rate after changing tasks and repeated blocks as far as we could hold the neuron. We used standard electrophysiological techniques to record single neuronal activity (Hasegawa et al. 1998b). Recording sites were histologically localized in the area anterior to the arcuate sulcus and medial to the principal sulcus (see Fig. 1B). No eye movement was elicited by intracortical microstimulation (100 µA) in this area, indicating that the recording area was outside of the frontal eye field. We generated rasters and average spike density histograms (binwidth, 10 ms) aligned on the cue onset. To detect significant changes in activity related to search, we compared cue-period activity with activity during fixation period (400 ms preceding cue presentation). Cue-period activity was defined as an activity in 200 or 300 ms starting 50, 100, 150, or 200 ms after the cue onset, which was automatically calculated to maximize activity. For the analysis on the close time course of population histograms, spike density function sampled at 1 kHz was convolved with a Gaussian filter (SD = 10 ms). All experiments were performed in accordance with the "Guidelines for the Care and Use of Laboratory Animals" of the National Institutes of Health (1985) and the "Guide for Care and Use of Laboratory Primates" published by Primate Research Institute of Kyoto University (1986).

View larger version (39K): [in this window] [in a new window]   Fig. 1. A: sequence of trial events of the delayed visual search task. The search array appears and then disappears while the fixation spot is lit. The monkey makes the saccade toward the remembered location soon after the fixation spot disappears. T and D represent the target and distractors, respectively. The arrow indicates a saccadic eye movement. B: schematic drawings of recording sites. The recorded sites (shaded area) were located anterior to the arcuate sulcus and medial to the principal sulcus. AS, arcuate sulcus; PS, principal sulcus. C: search-related activity of a single prefrontal (PF) neuron. Neural response was shown in each trial condition. Left and right columns, respectively, represent "bottom left" and "top right" target location. Top and bottom rows, respectively represent "fine" and "coarse" target objects. The top panel in each condition indicates the search array, which was presented for 0.5 s while the monkey maintained central fixation. The target object is the odd item among search array, which is different from distractors in spatial frequency of checkerboard pattern. The area surrounded by dashed line and also marked by RF in each stimulus panel represents a response field for the target object. In the raster displays, vertical tick marks and circles represent times of spikes and saccades. Neural responses are aligned on the cue onset. Left and right vertical lines through rasters indicate the onset of the search cue and its offset (beginning of the delay period).

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

We recorded the activity of 157 single cells from the dorsolateral PF cortex (Fig. 1B) in two rhesus monkeys. Of them, 53 (32 from monkey 1 and 21 from monkey 2) showed neuronal activity during the cue period. Figure 1C shows an example of a neuron that had a response field specific for a search target. The neuron responded to the onset of the search array when the monkey had to find the higher spatial target and it appeared at the bottom left (Fig. 1C, top left). After a transient response to the onset of the search array, the activity continued during a delay period. Activity in both the cue and delay period was significantly greater than the fixation period (Wilcoxon signed-rank test, P < 0.05). The neuron responded much less when a distractor appeared at the same place (Fig. 1C, top right). The identification of a stimulus as the target determined the response, not its spatial frequency. When the coarse spatial frequency stimulus was the target, the neuron responded to it (Fig. 1C, bottom left), although it did not respond to the stimulus when it was the distractor. It is also clear from Fig. 1C, bottom right, that these responses did not depend on the existence of the fine object in this target-specific response field. Of 53 visual neurons, 34 (64%) showed significant difference in cue-period activity between trials for target in the response field and distractors in the field (Mann-Whitney U test, P < 0.05). Of these 34, more than 1/2 of the neurons (n = 20, 61%) did not show a significantly higher cue-period activity to distractors presented in the response field than that of the fixation period (Wilcoxon signed-rank test, P > 0.05).

It could be that the increased response to the search target was merely due to its selection as a saccade target. We compared the activity in the DVS task to their activity in a simple ODR task (Funahashi et al. 1989), which had the same time course and set of stimuli as the DVS task. Figure 2A shows an example tested in both tasks. This neuron had selectivity for target location in both tasks; we estimated the response field to be a restricted area covering right and top right. However, there was difference in magnitude of cue-period activity between DVS and ODR tasks. The response to the search target in the DVS task was significantly higher than that to the detection target in the ODR (Mann-Whitney U test, P < 0.05). For 22 neurons tested, we compared the responses to the target in both tasks (Fig. 2B). Activity in the DVS task was greater for 19 neurons, and the population activity was also statistically higher (Wilcoxon signed-rank test, P < 0.001).

View larger version (28K): [in this window] [in a new window]   Fig. 2. A: quantitative analysis for a PF neuron selective to target location during a delayed visual search (DVS) and an oculomotor delayed response (ODR) task. Location of each histogram corresponds to the target location in search array. Neural responses are aligned on the cue onset (vertical dashed line). The cue was presented for 1 s in this recording session. The polar plots in both tasks show the average activity in cue-period (solid thick line) and control (fixation) period (dashed thin line) to 6 target locations. The area marked by RF and surrounded by a dotted line means the estimated response field for target location. B: comparison of DVS and ODR activities for 22 neurons tested.

Monkeys occasionally made errors in the DVS task, making saccades to a distractor rather than to the target location. In these error trials the neuron responded to the target as if it had been a distractor (Fig. 3A), failing to respond even though the unique stimulus was in the response field. For 23 neurons with enough error trials for this analysis, we compared magnitude of responses to target between correct and error trials (Fig. 3B). For 20/23 neurons the response to the target was less in the error trials than that in correct trials, and the population activity was also significantly less (Wilcoxon signed-rank test, P < 0.01).

View larger version (28K): [in this window] [in a new window]   Fig. 3. A: comparison of neuronal responses on 3 types of correct and error trials in the neuron illustrated in Fig. 2A; response to the target inside the response field on correct trials (left); response to the target outside the response field on correct trials (middle); and response to the target inside the response field on error trials (right). The response is much less when the target appears in the response field but the saccade is to a distractor. Abbreviations are the same as those in Figs. 1 and 2. B: comparison of averaged activities on correct and error trials for 23 neurons tested.

As shown in Figs. 1C and 2A, PF neurons showed a differential activity for the location of the search target during the cue period. Cue-period activity was often followed by delay- and/or go-period activities. To examine when the neurons selected the target and how the information could be used for the upcoming saccade, we made population histograms synchronized on cue onset and saccade onset for target and distractors in the response field (Fig. 4, top). Information about target location developed very early (starting about 135 ms after cue onset) and was maintained through the delay period. Presaccadic activation appeared just before the monkey made a saccadic eye movement toward the remembered location of the target. The presaccadic burst and beak were located at around 250 and 115 ms before the saccade onset, respectively. Figure 4, bottom, shows a comparison of time course of activity on DVS and ODR trials. As shown in Fig. 2, PF activities were enhanced in the search task compared to the detection trials. Search enhancement began about 180 ms after cue onset and lasted until 100 ms before the time of saccade, although the difference during the early part of the delay period was smaller.

View larger version (38K): [in this window] [in a new window]   Fig. 4. Preferred vs. nonpreferred location: population (average) activity of 24 PF neurons selective for the target location during the DVS task. Thick and thin lines represent activity for preferred and nonpreferred target locations, respectively. The difference in activity during every 500-ms epoch after cue onset trough to saccade onset (0-500, 500-1,000, 1,000-1,500, 1,500-2,000 ms from cue onset and 500-0 ms from saccade onset) between preferred and nonpreferred locations was significantly higher (t-test, P < 0.05) than the 500-ms epoch before cue onset. DVS vs. ODR task: population activity of 12 PF neurons selective for the target location tested in both DVS and ODR tasks. Thick and thin lines represent activity for preferred target locations on DVS and ODR trials, respectively. The difference in activity during 1st 500-ms epoch after cue onset (0-500 ms) and presaccadic epoch (500-0 ms from saccade onset) between DVS and ODR task was significantly higher (t-test, P < 0.05) than that before cue onset (500-0 ms).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

In these experiments we show that neurons in monkey PF cortex respond to stimuli in a complex array only when those stimuli dictate a behavioral choice. Neuronal activity related to visual search has been well examined in the frontal eye field (FEF) adjacent to the periprincipal sulcal area from which we recorded. The FEF is thought to be a cortical center of eye movement (Bruce and Goldberg 1985), and the visual response of FEF neurons to a peripheral stimulus is enhanced during a saccade task compared to a fixation task (Goldberg and Bushnell 1981). Furthermore, Burman and Segraves showed that FEF neurons were active in saccade planning while the animal scanned a natural image (Burman and Segraves 1994). Schall and his group (Schall et al. 1995; Thompson et al. 1996) studied the responses of FEF visuomovement neurons in a search task based on color or frequency difference. Those neurons gave identical transient responses to target and distractors and distinguished between target and distractor after the transient response at around 120-150 ms. Periprincipal neurons started distinguishing between target and distractors at around 135 ms, that is, at the same time of FEF neurons, yielding much smaller if any transient responses to distractors. In this regard they resemble the response of FEF visuomovement neurons that fail to respond to distractor stimuli when the monkey was overtrained on a simple color discrimination task (Bichot et al. 1996), although the latency of differential activity was earlier (at around 80 ms). The differential latency found in our study was consistent with previous studies in which a match-to-sample paradigm was used to test PF neurons [around 140 ms by Rainer et al. (1998) and 100-200 ms bin by Hasegawa et al. (1998b)]. While matching-to-sample is mainly guided by a memory representation, a pure odd-item search in which target object and location are shuffled randomly trial by trial may involve different mechanisms. We confirmed that PF neurons did contribute in early processing of visual search over the paradigm. Furthermore, we also found that a differential activity for target location was enhanced on search trials compared to that on detection trials. This search enhancement may depend on the level of focal attention. Unlike detection trials, subjects may require a higher level of focal attention to make sure of the target location and plan the appropriate choice, especially in a delayed condition. The enhancement, which was most evident 180 ms after the cue and again 250-100 ms before the saccade, may reflect the increase of focal attention and the contribution of signaling attentional movement, respectively.

Compared to the FEF neurons, periprincipal neurons do not seem tied to the outer environment. In general, their onset latency to visual stimulus is longer than that of FEF neurons, and in some cases they appear to care less about distractors in the response field. Their presaccadic burst suggests that they plan or signal directional saccades long before its execution, but the magnitude of activity could be modulated by the behavioral context. Further studies are necessary to make clear the functional differences of both areas. The response field of periprincipal neurons also differs from the response field in the other cortices. Feature-based target selection may occur in the inferotemporal (IT) cortex, which is selective to physical property of a target object (Chelazzi et al. 1998). Activity in this area may be a source of weak selectivity for a target object observed even in our study. It is possible that lateral intraparietal area (LIP), a part of the parietal cortex, is involved in finding an odd-item. LIP neurons represent salient objects in the response field in a complex visual scene even when stimulus is not a saccade target (Gottlieb et al. 1998). LIP may give information of the location of odd objects to the PF cortex. Thus the PF cortex may interact with these FEF, IT, and LIP areas to search for a saccade target among distractors.